Help, a problem about T85 running

[widesky_2005@yahoo_com_cn]

Hi, everyone

Excuse me, now I have a problem about running cam3 model on SGI.
I choose the T85 resolution, and my building process has been successfully finished.
Unfortunately, during my running process, the model suffers a ENDRUN after 10 steps.
In the file "output", the error is presented as follows:


NSTEP = 11 9.142942053303612E-05 1.470162251202247E-05 253.788 9.84682E+04 2.510360240790605E+01 2.41 0.94
nstep, te 12 3354539794.14015 -2.79927121380965
2.793938689495048E-004 98468.1675236412
CUBYDR: Departure point out of bounds: jcen,icount,nlon*plev= 36
6655 6656
ENDRUN: called without a message string


Then in the source code cubdyr.f90, I find that the model is stopped because the icount doesn't equal to the nlon*plev.

why? After all, I have successfully finished a T42 control simulation with different inputdata.

If somebody had same experience , I would like to receive your advice.

Many thanks in advance

 

[olson]

The "icount doesn't equal to the nlon*plev" message is a long-winded way of
saying you have exceeded the CFL condition (the message means that a "departure
point" in the semi-lagrangian calculation has gone beyond one grid box). Try
using a smaller timestep. Typically for T85, the timestep "DTIME" should be 600 seconds.

If that does not help, you might also try setting
DIVDAMPN = 1.

in your namelist. That will damp the divergence term for the first day of your
run. Hopefully that would give enough time for any initial shocks in the model
to settle down.